Milk quotas, based on an average fat content, severely limit milk production on UK farms. Predictions of the time-course of lactation are incorporated into most computerized herd management programs but these models take no account of food inputs, body weight change or milk composition. Dynamic models are generally used to simulate metabolic pathways and, as such, have little direct relevance to commercial milk production. Dynamic models can be converted to an adaptive-predictive model that partitions food energy into milk and non-milk constituents. This paper reports the development of an adaptive-predictive model to partition food into milk and non-milk components. Additional functions further partition milk energy into the principal constituents, fat, protein and lactose.